Merge pull request 'Implemented system metrics' (#472) from jpds/garage:system-metrics into main

Reviewed-on: Deuxfleurs/garage#472
Reviewed-by: Alex <alex@adnab.me>

src/block/manager.rs

@@ -129,8 +129,12 @@ impl BlockManager {
 			.netapp
 			.endpoint("garage_block/manager.rs/Rpc".to_string());
 
-		let metrics =
+		let metrics = BlockManagerMetrics::new(
-			BlockManagerMetrics::new(rc.rc.clone(), resync.queue.clone(), resync.errors.clone());
+			compression_level,
+			rc.rc.clone(),
+			resync.queue.clone(),
+			resync.errors.clone(),
+		);
 
 		let scrub_persister = PersisterShared::new(&system.metadata_dir, "scrub_info");
 

src/block/metrics.rs

@@ -5,6 +5,7 @@ use garage_db::counted_tree_hack::CountedTree;
 
 /// TableMetrics reference all counter used for metrics
 pub struct BlockManagerMetrics {
+	pub(crate) _compression_level: ValueObserver<u64>,
 	pub(crate) _rc_size: ValueObserver<u64>,
 	pub(crate) _resync_queue_len: ValueObserver<u64>,
 	pub(crate) _resync_errored_blocks: ValueObserver<u64>,
@@ -25,9 +26,23 @@ pub struct BlockManagerMetrics {
 }
 
 impl BlockManagerMetrics {
-	pub fn new(rc_tree: db::Tree, resync_queue: CountedTree, resync_errors: CountedTree) -> Self {
+	pub fn new(
+		compression_level: Option<i32>,
+		rc_tree: db::Tree,
+		resync_queue: CountedTree,
+		resync_errors: CountedTree,
+	) -> Self {
 		let meter = global::meter("garage_model/block");
 		Self {
+			_compression_level: meter
+				.u64_value_observer("block.compression_level", move |observer| {
+					match compression_level {
+						Some(v) => observer.observe(v as u64, &[]),
+						None => observer.observe(0 as u64, &[]),
+					}
+				})
+				.with_description("Garage compression level for node")
+				.init(),
 			_rc_size: meter
 				.u64_value_observer("block.rc_size", move |observer| {
 					if let Ok(Some(v)) = rc_tree.fast_len() {

src/rpc/lib.rs

@@ -17,3 +17,5 @@ mod metrics;
 pub mod rpc_helper;
 
 pub use rpc_helper::*;
+
+pub mod system_metrics;

src/rpc/system.rs

@@ -38,6 +38,9 @@ use crate::replication_mode::*;
 use crate::ring::*;
 use crate::rpc_helper::*;
 
+#[cfg(feature = "metrics")]
+use crate::system_metrics::*;
+
 const DISCOVERY_INTERVAL: Duration = Duration::from_secs(60);
 const STATUS_EXCHANGE_INTERVAL: Duration = Duration::from_secs(10);
 
@@ -103,6 +106,8 @@ pub struct System {
 	consul_discovery: Option<ConsulDiscovery>,
 	#[cfg(feature = "kubernetes-discovery")]
 	kubernetes_discovery: Option<KubernetesDiscoveryConfig>,
+	#[cfg(feature = "metrics")]
+	metrics: SystemMetrics,
 
 	replication_mode: ReplicationMode,
 	replication_factor: usize,
@@ -275,6 +280,9 @@ impl System {
 			cluster_layout_staging_hash: cluster_layout.staging_hash,
 		};
 
+		#[cfg(feature = "metrics")]
+		let metrics = SystemMetrics::new(replication_factor);
+
 		let ring = Ring::new(cluster_layout, replication_factor);
 		let (update_ring, ring) = watch::channel(Arc::new(ring));
 
@@ -365,6 +373,8 @@ impl System {
 			consul_discovery,
 			#[cfg(feature = "kubernetes-discovery")]
 			kubernetes_discovery: config.kubernetes_discovery.clone(),
+			#[cfg(feature = "metrics")]
+			metrics,
 
 			ring,
 			update_ring: Mutex::new(update_ring),

src/rpc/system_metrics.rs

@@ -0,0 +1,33 @@
+use opentelemetry::{global, metrics::*, KeyValue};
+
+/// TableMetrics reference all counter used for metrics
+pub struct SystemMetrics {
+	pub(crate) _garage_build_info: ValueObserver<u64>,
+	pub(crate) _replication_factor: ValueObserver<u64>,
+}
+
+impl SystemMetrics {
+	pub fn new(replication_factor: usize) -> Self {
+		let meter = global::meter("garage_system");
+		Self {
+			_garage_build_info: meter
+				.u64_value_observer("garage_build_info", move |observer| {
+					observer.observe(
+						1,
+						&[KeyValue::new(
+							"version",
+							garage_util::version::garage_version(),
+						)],
+					)
+				})
+				.with_description("Garage build info")
+				.init(),
+			_replication_factor: meter
+				.u64_value_observer("garage_replication_factor", move |observer| {
+					observer.observe(replication_factor as u64, &[])
+				})
+				.with_description("Garage replication factor setting")
+				.init(),
+		}
+	}
+}